Neurobiological Correlates of Social Conformity and Independence During Mental Rotation PDF

Neurobiological Correlates of Social Conformity and Independence During Mental Rotation Gregory S. Berns, Jonathan Chappelow, Caroline F. Zink, Giuseppe Pagnoni, Megan E. Martin-Skurski, and Jim Richards Background: When individual judgment conflicts with a group, the individual will often conform h...

Neurobiological Correlates of Social Conformity and Independence During Mental Rotation Gregory S. Berns, Jonathan Chappelow, Caroline F. Zink, Giuseppe Pagnoni, Megan E. Martin-Skurski, and Jim Richards Background: When individual judgment conflicts with a group, the individual will often conform his judgment to that of the group. Conformity might arise at an executive level of decision making, or it might arise because the social setting alters the individual’s perception of the world. Methods: We used functional magnetic resonance imaging and a task of mental rotation in the context of peer pressure to investigate the neural basis of individualistic and conforming behavior in the face of wrong information. Results: Conformity was associated with functional changes in an occipital–parietal network, especially when the wrong information originated from other people. Independence was associated with increased amygdala and caudate activity, findings consistent with the assumptions of social norm theory about the behavioral saliency of standing alone. Conclusions: These findings provide the first biological evidence for the involvement of perceptual and emotional processes during social conformity. Key Words: Social conformity, fMRI, perception, mental rotation, whether people capitulate to the group despite knowing that the parietal lobe group is wrong (i.e., an executive decision-making process) or whether they conform because the group has altered their Man is born free; and everywhere he is in chains. perception. An individual’s judgment has been assumed to reflect what that individual perceives (Coren and Enss 1993), but other Jean Jacques Rousseau, The Social Contract (1762) factors can influence perception, such as categorization (Gold- I ndividuals in democratic societies are free to make choices stone 1995) and social class (Stapel and Koomen 1997), indicat- and express their opinions, but the price of such freedom is ing that perception is not simply a passive process. Although sometimes the subjugation of individual choice to the gen- many exogenous factors can modify judgments, the mechanism eral will—Rousseau’s social contract. The accepted resolution of by which this modification occurs and the degree to which these the conflict between individual and group decision making is the modifications are based on perception or decision making well-known “rule of the majority.” There is a sound basis for this remains unknown. compromise: a group of individuals is statistically more likely to With a modification of the Asch paradigm, we used fMRI to make a better decision than any one person alone (Arrow 1963; examine the alterations in brain activity associated with social Grofman and Feld 1988). But the superiority of the group conformity and independence. We hypothesized that if social disappears when individuals influence each other (Ladha 1992). conformity resulted from conscious decision making, this would Moreover, individuals might capitulate to a group, not as part of be associated with functional changes in prefrontal cortex, the social contract, but because the unpleasantness of standing whereas if social conformity was more perceptually based, then alone makes the majority opinion more appealing than one’s activity changes would be seen in occipital and parietal regions. own beliefs (Cialdini and Goldstein 2004). How and why this In the present study, instead of judging relative line lengths, happens has been debated contentiously. Here, we bring func- normal adult volunteers mentally rotated three-dimensional ob- tional magnetic resonance imaging (fMRI) to bear on the prob- jects and had to judge whether the objects were the “same” or lem of social conformity. “different” (Shepard and Metzler 1971). This task has been well The modification of an individual’s judgment under the characterized both behaviorally and with functional brain imag- pressure of a group was first demonstrated in a series of ing, and activity in the intraparietal sulcus increases during the experiments by Solomon Asch (Asch 1951, 1952). In these process of mental rotation (Alivisatos and Petrides 1997; Cohen experiments, participants were presented with line segments and et al 1996; Jordan et al 2001; Tagaris 1997). To induce conformity made judgments about their relative lengths. Although these while performing the mental rotation task, participants were tasks were perceptually simple, participants frequently gave the presented with the responses of four peers, who, unknown to wrong answers when a group of peers was also giving the wrong the participant, were actors giving wrong answers half of the answer. These findings raised the fundamental question of time. To differentiate the effect of social conformity from the conflict engendered by misinformation, each participant per- formed one round of trials with the group and another round in From the Department of Psychiatry and Behavioral Sciences (GSB, JC, CFZ, which the actors were replaced by computers. GP, MEM-S), Emory University School of Medicine; Wallace H. Coulter Department of Biomedical Engineering (GSB), Georgia Institute of Tech- nology and Emory University School of Medicine; and Richards, LLC (JR), Methods and Materials Atlanta, Georgia. Address reprint requests to Gregory S. Berns, M.D., Ph.D., Emory University Participants School of Medicine, Department of Psychiatry and Behavioral Sciences, Participants were 33 normal, right-handed volunteers (14 1639 Pierce Drive, Suite 4000, Atlanta, GA 30322; E-mail: gberns@ female, 19 male), with a mean (SD) age of 26.0 (5.8) years (range, emory.edu. 19 – 41 years). An additional two participants were studied but Received September 30, 2004; revised February 1, 2005; accepted April 5, were excluded from the analysis because debriefing indicated 2005. that they did not understand the task. One participant had an 0006-3223/05/$30.00 BIOL PSYCHIATRY 2005;58:245–253 doi:10.1016/j.biopsych.2005.04.012 © 2005 Society of Biological Psychiatry 246 BIOL PSYCHIATRY 2005;58:245–253 G.S. Berns et al artifact on her functional brain images and was discarded from tems, Albany, California). Responses were registered on a fiber the analysis, leaving a total of 32 participants. All participants optic button box, and the RTs were logged through Presentation. gave written informed consent to a protocol approved by the Each trial began with a group decision phase of variable Emory University Institutional Review Board. In addition to the duration. The participant was told that during this period, the participants, four actors (two female, two male) were hired to group would see the stimuli, and after all of the group members play the role of the group. The actors were selected to be had made their decision about sameness, their responses would representative of the age, gender, and ethnicity of the people that be displayed (hence the variable duration). In one third of the historically have volunteered for our imaging studies. trials, the group’s response was hidden from the participant, and this was denoted by an “X” next to each person’s icon (baseline condition). By predesign, in one third of the trials the group was Procedure unanimously wrong (split 50-50 between Same and Different Each participant was escorted to the MRI waiting room. For trials), and in the remaining third, the group was correct (also each participant, the actors arrived in haphazard fashion, osten- evenly distributed between Same and Different trials). Sixteen sibly as participants in the same experiment. When the partici- trials of each type (baseline, group correct, group incorrect) were pant and all of the actors were present, the group was escorted ordered randomly. An additional six “split-decision” trials were to the scanner console room, where the task was explained and randomly interspersed throughout each round. In these trials, the the consent forms signed by everyone. We used a digital camera group split either 2-2 or 3-1 on the same– different judgment, and to capture images of the participant and actors. The images were these trials were used solely to maintain believability about the then cropped and used in the subsequent task. interaction. These trials were not used in the analysis. To form group cohesiveness, the participant and the actors After the group decision phase, the participant was presented simultaneously played a practice round of 20 trials on a network with the group’s response for 3 sec (to guarantee that they saw of five laptop computers set up in the scanner control room. All it), and then the participant was presented with the stimulus pair. of the group actors participated in this practice simultaneously The mental rotation period ended when the participant indicated with the participant, and the procedure was repeated for every their choice by a button press. There was a 12-sec time limit to participant as though the actors were also appearing naively for respond during the mental rotation period. The percent of trials the experiment. This included taking pictures of the people each in which the participant did not answer in time was small (mean time and signing consent forms each time as well. The practice 2.0%, SEM 0.5%) across the entire experiment. After the re- round itself was constructed of trials with no group feedback. sponse, the participant’s and the group’s responses were dis- There were two reasons for this. First, we wanted to be sure that played for 3 sec. the participant was able to do the task on their own, without Each participant performed two rounds as described above: relying on the group. Second, we wanted to avoid biasing the once with the group and once with the group replaced by participant in any way regarding the group’s intentions before computers. The order was counterbalanced across participants scanning. After the practice round, the participant was placed in the and gender. The instructions for the computer round were scanner, and the actors were dismissed until the next participant purposely vague so as not to bias a particular belief about a arrived. computer’s validity: “Based on a simple algorithm, the computer will make a determination of whether the objects are the same or Experimental Task different.” The participant was presented with pairs of abstract three- dimensional stimuli from the Shepard and Metzler set (Figure 1) Imaging (Shepard and Metzler 1971). Participants knew that their re- Imaging was performed on a Siemens 3 Tesla Trio whole- sponses would be visible to the other people. Half of the body MRI scanner (Siemens AG, Erlangen, Germany). Head stimulus pairs represented the same object (hence called Same) movement was restricted with padding. After performing auto- rotated relative to each other, and the other half of the stimulus matic shimming and acquiring a scout image, we performed two pairs represented objects that were mirror images of each other runs of echo-planar imaging (EPI) to maximize the blood oxy- (hence called Different) and also rotated relative to each other. genation level– dependent (BOLD) effect associated with neuro- The task was to judge whether the objects could be rotated to nal activation (repetition time ! 2.345 sec, echo time ! 30 msec, match each other and hence be judged as being Same, or not, in 64 " 64 matrix, 35 slices of 3-mm cubic voxels). The number of which case they would be judged as Different. The selection of scans varied in each run, depending on the speed of the stimuli was based on the original Shepard and Metzler data participant, but ranged from 363 to 551. showing a linear increase in reaction time (RT) with angle of disparity. We used stimulus pairs with angles of disparity ranging Debriefing Procedure from 100° to 180° (mean 134°), to yield moderately difficult After the scan procedure, participants completed a question- stimuli and to maintain believability around the group’s wrong naire to assess their perceptions about the experiment. Using a answers. The direction of rotation was randomized and counter- visual analogue scale (VAS), participants were asked to rate their balanced across conditions. Because disparity cannot be defined own accuracy, the accuracy of the other players, and the uniquely for mirror image pairs, we approximated the degree of accuracy of the computers. Participants also checked a series of difficulty in the mirror image set by using a set of stimuli that Yes/No questions: 1) Did you ever follow the answer of the other roughly matched the RTs of the Same stimuli. Importantly, all of players? If yes, because (select all that apply): a) you were sure the experimental manipulations, namely whether the group or you had the right answer and that the majority response was computers were correct or incorrect, were balanced between right; b) you were sure you had the right answer, but picked the Same and Different stimuli pairs. Stimulus presentation and wrong answer to go with the majority response (explain why); c) response acquisition were coordinated by a laptop computer you were not sure about your answer and decided to go with the running the software Presentation 0.55 (Neurobehavioral Sys- majority response (explain why). Similar sets of questions were www.sobp.org/journal G.S. Berns et al BIOL PSYCHIATRY 2005;58:245–253 247 Figure 1. Participants were presented with pairs of three-dimensional objects on a computer screen during a mental rotation period, and they had to decide whether the objects were the “same” (can be rotated to match) or “different” (no rotation can make them match). To induce social conformity, each trial began with the objects being shown first to a group of peers (Group; top panel). In actuality, the group was composed of actors, and their re- sponses were predetermined. After a variable-dura- tion decision phase, the collective response of the group was displayed to the participant. This ensured that the participant would see the group’s response. After 3 sec, the same pair of objects was displayed to the participant. In the example shown, the objects are different, but the group has unanimously said they are the same (the participant has not re- sponded yet). The participant responded with a but- ton press, indicating whether the objects were the same or different. Trial types were randomized across three conditions: group correct, group incor- rect (as shown), and baseline (responses blinded to participant with an “X”; bottom panel). One run of 48 trials was performed with the group, and another run of the same 48 trials was performed with the group replaced by computers (bottom panel), in which the faces of the group were substituted with computer icons. The order of group and computer runs was counterbalanced across participants and gender. asked for the cases of going along with the computers, going each aimed at contrasting a specific main effect. This included against the group, and going against the computers. All but three separate models based on external information (correct, incor- participants filled out the questionnaire. rect, or none), stimulus type (same or different), participant’s response (correct or incorrect), as well as a model that catego- Image Analysis rized trials on the basis of a combination of these effects. The All image analysis was performed with SPM2 (Wellcome Depart- more fine-grained categorization of trial types meant a concom- ment of Imaging Neuroscience, London, United Kingdom). Func- itant loss of observations for some participants who did not make tional images were first interpolated to correct for slice timing. many errors, so this model was used only in the analysis of Functional images were then corrected for motion and then conformity and independence (see Supplement 1 for effects and spatially normalized to the SPM2 EPI template reslicing at 3 " 3 their durations). Contrasts were generated from the design matrix " 3-mm resolution and smoothed with an 8-mm isotropic at the individual participant level and then entered into a gaussian kernel (Ashburner and Friston 1999; Friston et al 1995). second-level analysis for statistical inference. We considered Statistical analysis was performed with a two-step procedure, significant activation to have occurred in a cluster of at least 5 first at the individual level and then at the cohort level, providing voxels (except as noted otherwise) if the peak activation level statistical inference in a random-effects model across participants was significant at p #.001 (uncorrected for multiple compari- (Friston et al 1999). Several different first-level models were used, sons). www.sobp.org/journal 248 BIOL PSYCHIATRY 2005;58:245–253 G.S. Berns et al was assessed by the correlation between self-rated accuracy (Self) and actual overall accuracy (Actual). Linear regression showed a strong correlation between these two measures where Self !.17 $.69 " Actual (R !.67, p #.0001, df ! 27). If we assume an intercept at the origin, then Self !.90 " Actual, indicating that participants, on average, slightly underestimated their true accuracy but were within 10% of their true value. On average, the group’s accuracy was rated as 65.1% (SEM 2.5%), and the computers’ accuracy was rated as 61.2% (SEM 2.7%). These were not significantly different from each other [two-tailed paired t (28) ! 1.491, p !.147], indicating that the participants did not view the group or computers as more reliable than one another. When asked why they went along with the group, 82.8% of participants indicated that on some trials they were sure that they were right and that the group was right too (89.7% for the computer), 58.6% indicated that on some trials Figure 2. Mean error rates as a function of the source (Computers or Group) they were not sure about their answer and decided to go with the and type (Correct or Incorrect) of information. The actual number of errors in majority (44.8% for the computer), and 3.4% indicated that were the Computers and Group conditions are shown on the right-hand scale sure they had the right answer but went with the majority (the number of errors in the baseline condition must be multiplied by two, response anyway (3.4% for the computer). because there were twice as many trials in this condition). Repeated-mea- sures analysis of variance revealed a large main effect of the type of infor- mation (Correct, Incorrect, or None) on error rates [F(2,31) ! 31.29, p #.0001], and post hoc comparisons indicated that incorrect information re- sulted in significantly greater error rates (p #.0001) than both baseline (green) and correct information (which were not significantly different from each other, p !.348). The interaction of the source of information (Group or Computers) with the type of information (Correct, Incorrect, or None) was also significant [F(2,31) ! 6.53, p #.004], and the post hoc comparison indicated that the error rate when the group gave wrong information (mean 41%, SEM 5%) was significantly greater than when the computers gave wrong information (mean 32%, SEM 4%) [***paired t(32) ! 3.55, p #.001]. Results Behavioral Measures of Conformity Conformity was defined as agreeing with the exogenous source of information, either peers or computers, when the information was wrong. Conformity was measured behaviorally by the change in error rates of the participants between their baseline performance and the conditions in which exogenous information was presented (Figure 2). The baseline error rate was computed for each participant from the trials in which no group (or computer) information was given (mean 13.8%, SEM 2%). The error rate increased to 41% (SEM 5%) when the group gave wrong information, which was significantly greater than when the computers gave wrong information (mean 32%, SEM 4%) [paired t (32) ! 3.55, p #.001]. There were significant differences in RTs, and these differ- ences depended on several factors (Figure 3). After adjusting for Figure 3. There were significant differences in reaction times (RTs), and the effect of Same/Different stimuli, there was a significant these differences depended on several factors. A mixed-model analysis with lengthening of RT when external information was present repeated effects showed that the stimulus pair (Same or Different) had a [F (2,579) ! 20.27, p #.0001]. But restricting the analysis to trials significant effect on RT [F(1,579) ! 66.38, p #.0001], and a post hoc com- in which incorrect information was provided, there was no parison indicated that Different stimuli took.815 sec (SEM.100 sec) longer significant difference in RT between going with (i.e., Conformity) than Same stimuli. After adjusting for the effect of Same/Different stimuli in this model, there was a significant lengthening of RT when external infor- or against (i.e., Independence) the information [F (1,190) ! 2.65, mation was present [F(2,579) ! 20.27, p #.0001]. Moreover, participants p !.105], indicating that participants did not take longer for one took, on average, 1.11 sec (SEM.18 sec) longer to give an incorrect response behavior or the other. Notably for the subsequent image analysis, than a correct response [t(32) ! 6.2, p #.0001], but this was modulated by the source of the external information, either Group or Comput- an interaction with the nature of the external information presented ers, also did not have a significant effect on RT [F (1,576) !.554, [F(2,579) ! 20.87, p #.0001]. Restricting the analysis to trials in which p !.457]. incorrect information was provided, there was no significant difference in RT between going with (i.e., Conformity) or against (i.e., Independence) the information [F(1,190) ! 2.65, p !.105], indicating that participants did not Debriefing Questionnaire take longer for one behavior or the other. Notably for the subsequent image Participants were debriefed through a combination of a VAS and analysis, the source of the external information, either Group or Computer, specific questions. The reliability of the VAS ratings for accuracy also did not have a significant effect on RT [F(1,576) ! 0.554, p !.457]. www.sobp.org/journal G.S. Berns et al BIOL PSYCHIATRY 2005;58:245–253 249 Figure 4. Effects of different forms of external information on brain activation during mental rotation. At baseline (mental rotation with no external information), a wide network of both cortical and subcortical regions was activated (inset and red regions) (all overlays thresholded at p #.001, uncorrected for multiple comparisons, and cluster size % 5 voxels). Montreal Neurological Institute coordinates of selected axial slices are indicated adjacent to each slice. External information, regardless of the source, was associated with decreased activation in a small subset of this mental rotation network (dark blue), primarily in frontal regions and the supplementary motor area (slices at $55 and $70 mm). In contrast, external information was associated with increased activation (light blue) in regions largely outside of the mental rotation network, particularly in the right supramarginal gyrus (slices at $31 and 34 mm). When the effects of external information were compared according to the source (Group % Computers), only two small clusters adjacent to the mental rotation network in the right intraparietal sulcus were identified (green). Moreover, the activity in these clusters was largely accounted for by the subset of trials in which the participants conformed to incorrect information from the Group (yellow). An additional two clusters in the occipital cortex were identified as differentiating between conforming to the Group versus the Computers (slice at $8 mm). Thus, the main effect of social conformity (i.e., conforming to the Group relative to conforming to the Computers) was exhibited within the most posterior aspects of the baseline mental rotation network. Because of the anatomic confluence of these different effects within the right intraparietal sulcus, a region-of-interest analysis was performed on this area (circle). Brain Responses During Mental Rotation consistent with previous brain imaging studies of mental rotation All of the brain imaging results refer to changes in activity (Alivisatos and Petrides 1997; Cohen et al 1996; Jordan et al 2001; during the period of mental rotation, which was defined as the Tagaris 1997), and because of the large extent of activations, we period beginning with the presentation of the stimulus pair and do not report specific cluster locations. ending with the participant’s judgment, excluding the actual response (modeled separately). Effects of External Information We used the trials in which participants received no external Two distinct effects of external information were observed, information to delineate the baseline mental rotation network. regardless of the source of information. First, the presence of This network provided an anatomic reference against which the external information was associated with decreased activation in effects of external information could be gauged. Mental rotation a subset of the baseline mental rotation network (Figure 4, dark was associated with increases in BOLD activity in a network of blue). These activations were positive but of a decreased mag- brain regions including occipital cortex, parietal cortex, and parts nitude from baseline mental rotation. With the exception of a of the frontal cortex (Figure 4, inset). This network was broadly single cluster in the left superior parietal lobe, all of these clusters www.sobp.org/journal 250 BIOL PSYCHIATRY 2005;58:245–253 G.S. Berns et al Table 1. Differential Activations During Mental Rotation, Where Group % Computers, and the External Information Was Incorrect MNI Coordinates Brain Regiona (x, y, z) Cluster Size t Statistic Social Conformity (Participant Incorrect) L superior occipital gyrus &18, &90, 9 21 4.99 R superior occipital gyrus 24, &87, 9 34 4.87 R intraparietal sulcus 24, &69, 36 21 4.83 R intraparietal sulcus 18, &57, 48 20 4.09 Social Independence (Participant Correct) R amygdala 15, &3, &18 3b 4.30 R caudate head 6, 12, 3 7b 3.85 Voxels were selected for p #.001 (uncorrected) and extent ! 10 voxels (except as noted). MNI, Montreal Neurological Institute; L, left; R, right. a Brain regions determined from the Duvernoy atlas (Duvernoy 1999). b Because of the small size of these structures, the extent threshold was relaxed. were located frontally. Second, external information was associ- and whether the participant gave a correct or incorrect response ated with increased activation (Supplement 1) in regions largely (i.e., social conformity) (Figure 5). outside, but in close proximity to the mental rotation network In contrast to social conformity, social independence resulted (Figure 4, light blue). The right supramarginal gyrus was the in only two small clusters of activation: in the right amygdala and largest area exhibiting this increase in activation. head of the right caudate nucleus (Figure 6, Table 1). Differential Effects of Group and Computers Discussion To focus on the social component of information during We are interested here in the potency of social pressure in mental rotation, we first contrasted trials with Group Information inducing conformity and how information that originates from versus Computer Information, irrespective of whether the infor- humans, versus inanimate sources, alters either perception or mation was correct. Because we were interested specifically in decision making and the neural basis for such changes. When the social component, we identified those regions in which the activity was greater for information originating from the Group (i.e. Group Information % Computer Information). At a signifi- cance threshold of p #.001 (uncorrected) and extent ! 10, only one cluster was identified in the right intraparietal sulcus (Mon- treal Neurological Institute [MNI] coordinates: 24, &69, 39; t ! 4.52, cluster size ! 12 voxels) (Figure 4, green). To focus specifically on the social component, we formed contrasts from those trials in which the external information was incorrect and examined the differential effects of Group and Computers. Conformity was defined as those trials in which the participant went along with the external information when the information was incorrect, and independence was defined as those trials in which the participant gave the correct answer when the external information was incorrect. The difference between Group and Computers on these two subsets of trials therefore isolated the social element. Regions in which Group Conformity resulted in greater activation than Computer Confor- mity are shown in Table 1 and Figure 4 (yellow). Significant clusters were observed bilaterally in the visual cortex that overlapped the baseline mental rotation network. Two other clusters were identified in the right intraparietal sulcus that were at the anatomic confluence of the mental rotation network and the network identified as being more active to external informa- tion (Figure 4, circle). Figure 5. Region-of-interest (ROI) analysis of the right intraparietal sulcus. A Because the right intraparietal sulcus figured prominently in 12-mm diameter sphere was centered on the anatomic confluence of ef- several contrasts (e.g., baseline mental rotation, Group Info % fects shown in Figure 4 (Montreal Neurological Institute coordinates: 25, Computer Info, and Group Conformity % Computer Conformity), &66, 32). The best-fitting model by Akaike’s Information Criterion (Akaike and because the specific part of the sulcus implicated in social 1974) showed that there was a main effect of source (Group vs. Computers) conformity seemed to lie at the anatomic junction of external [F(1,631) ! 6.56, p !.01] and that there was a significant interaction be- information and mental rotation, we performed a region of tween the source of information and whether the participant gave a correct or incorrect response (i.e., Social Conformity) [F(1,631) ! 5.80, p !.016]. Post interest (ROI) analysis on this area. An ROI analysis allowed a hoc comparisons indicated that conforming to incorrect information from more detailed investigation of the effects driving the response in the Group was associated with significantly greater activity than conform- this area, showing that there was a main effect of source (Group ing to incorrect information from the Computers [***mean difference.271 vs. Computers) and an interaction between source of information (SEM.078), F(1,71) ! 11.98, p #.001]. www.sobp.org/journal G.S. Berns et al BIOL PSYCHIATRY 2005;58:245–253 251 conformity effect adds further evidence to the possibility that conformity was effected through a change in perception. Previ- ous studies have shown that activity in visual cortical neurons correlates with perception (Britten et al 1992), and in humans, visual cortex activity correlates more closely with perception than actual stimulus properties (Ress and Heeger 2003), a finding consistent with the results reported here. We found evidence for modulation of the same visual cortical regions when participants were about to conform to the group. The lack of concomitant activity changes in more frontal areas was highly suggestive of a process based, at least partially, in perception. Of course, changes in frontal activity could have occurred below our detection threshold, but with 32 participants, we think this unlikely. It is well known that attention can exert a “top-down” modulation of visuospatial processing (Pessoa et al 2003), and this might underlie the activity changes that we observed. When confronted with a situation in which the group was giving wrong information about a stimulus pair, it was possible that the wrong information evoked extra visual attention. If participants at- tended more strongly to humans than computers, then one would expect to see greater activation in attentional networks upon presentation of human information, even before the period of mental rotation. But when we compared the activation to Group and Computers before the appearance of the shapes, no differences in activity were observed in either the occipital or parietal cortex, even after dropping the threshold to p #.01. Although human information was more potent in inducing conformity and more strongly associated with changes in per- ceptual processing in parietal cortex, we cannot conclude that these effects were mediated simply because the participants attended more strongly to human faces. Another possibility is that the Group evoked a “virtual per- cept” that competed with that of the participant. Competing visual stimuli generally decrease visual cortex activation, but internally generated attention increases visual and parietal cortex activity (Kastner et al 1999). Trials in which a participant was about to conform represented precisely those trials in which the participant relied on the Group (or Computers). We can assume this information was both represented and processed in a manner substantially different from the shapes themselves, be- cause of the increase in activity outside the baseline mental Figure 6. Brain regions associated with greater activity during mental rota- rotation network (Figure 4, light blue). Even so, the processing of tion when participants went against the Group, relative to going against the this external information seemed to overlap the mental rotation Computers. In the subset of trials in which the external information was network in both the intraparietal sulcus and visual cortex (Figure incorrect, only the right amygdala (A) and the right caudate (B) showed 4, yellow). When the Group (or Computers) gave the wrong greater activity when Group Independence was compared with Computers Independence. Peak activations in these two regions were significant at p # answer, a “virtual” percept was created, and although partici-.001, uncorrected, but the extent of activation is shown thresholded at p # pants were always presented with one stimulus pair, the stimulus.01. See Table 1. could have represented two judgments. Visual cortex activity has been found to correlate with the degree of competition between participants conformed to the judgments of a group of peers, different percepts (Rees and Frith 1998), and here, we interpret relative to nonhuman sources, activity within the brain network the level of visual cortex activity to indicate the amount of that normally accomplishes the task of mental rotation was competition between the individual’s percept and the external altered. These findings indicate that with mental rotation, the percept provided by the Group (or Computers). The increase in effects of social conformity are exerted on the very same brain activity within the intraparietal sulcus suggested that the Group’s regions that perform the task. Although an extensive network of percept was stronger than that of the Computers. The intrapari- regions was involved in mental rotation, it was striking that the etal sulcus has been implicated in decision making (Shadlen and effects of social conformity were detected only in the most Newsome 2001), but the right intraparietal sulcus also orients posterior aspects—the occipital and parietal lobes. attention to extrapersonal space (Nobre et al 2004). The conflu- ence we found of external information with mental rotation in Perception Versus Decision Making this area suggests that this region might serve as a mixing ground Asch himself raised the possibility that social pressure could for internal and external percepts. When activity increased alter perception (Asch 1952). The posterior distribution of the beyond a critical threshold in the intraparietal sulcus, this might www.sobp.org/journal 252 BIOL PSYCHIATRY 2005;58:245–253 G.S. Berns et al have represented the “winning” of the external percept over the (Cialdini and Goldstein 2004). Although the peak activation participant’s own percept, and conformity became imminent. difference was highly significant, the spatial extent was small, Of course, this does not rule out an executive decision- and so we must view this as a tentative finding. Nevertheless, making process. The debriefing questionnaire partially ad- several fMRI studies of amygdala activation have reported small dressed the degree to which participants were consciously aware spatial extents (Phan et al 2004; Phelps et al 2001; Whalen et al of their judgments. Participants were surprisingly accurate in 1998). gauging their own performance as well as that of the Group and Another key component of social norm theory is that the the Computers. They tended to underestimate their own perfor- information provided by the group must be salient to the mance slightly and overestimate that of the Group and the individual. Besides the amygdala, the only other brain structure Computers, but there was no evidence that participants found differentially activated by social independence was the right either the Group or Computers more reliable than each other. caudate nucleus (Figure 6). Several studies in both nonprimates The vast majority of participants indicated that, at least on some and humans have recently implicated this region of the striatum trials, they went along with the external information because as critically involved in the processing of stimulus saliency they thought that they had arrived serendipitously at the same (Horvitz 2000; Zink et al 2003, 2004). The striatum seems to be correct answer. particularly involved in the selection of behaviors based on One possibility is that the external information created con- stimulus salience (Redgrave et al 1999). It is therefore not fusion in the participants’ minds where none existed before. surprising that the caudate should be most active when the Undoubtedly this contributed to some of the effect, but because participant behaved independently of the group, for this is the social conformity was defined as the difference between con- condition of most salience: conflicting information and social forming to a group and conforming to computers, an added isolation. attribute of information originating from humans must have been involved. Another possibility is that external information relieved Limitations some of the load on the participants’ mental rotation network. In Although we attempted to control for as many aspects of fact, we did find evidence for this. We observed decreased conformity as possible, there are limitations to our study. First, activations during mental rotation in frontal regions when exter- we had to modify Asch’s original experiment. Asch used a nal information was present, regardless of the source (Figure 4, perceptually unambiguous task of judging relative line lengths, dark blue). Although these regions tended to be restricted and so any errors participants made would have been caused by spatially, their distribution was markedly different than the social conformity. But even the most conformist of participants in aforementioned conformity changes. With the exception of one Asch’s experiments only went along with the group a few times. cluster in the superior parietal gyrus, all of these regions of In a neuroimaging study, this would not provide a sufficient decreased activation were anterior to the central sulcus. This number of trials to examine. Even with the mental rotation suggests that participants were still mentally rotating the objects, paradigm, there was a relatively low number of conformity trials, but the external information relieved the processing load at the averaging five to six per person. This necessitated scanning a output stage. Moreover, this “off-loading” effect did not depend large number of participants, which, although partially compen- on the source of information. sating for the low statistical power, might have led to type II statistical errors, especially with regard to the lack of frontal The Pain of Independence changes. Second, our experimental setup with a group of peers Compared with behavioral research of conformity, compara- might have created a “demand” effect on the participants, in tively little is known about the mechanisms of nonconformity, or effect inducing them to conform more than they would have independence. In one psychological framework, the group pro- under other circumstances. Finally, because of the visual nature vides a normative influence on the individual. Depending on the of the task, the conformity-associated changes in visual and particular situation, the group’s influence might be purely infor- parietal cortex might be specific to visual tasks. These limitations, mational—providing information to an individual who is unsure however, point the way toward future experiments—for exam- of what to do. More interesting is the case in which the individual ple, the role of automaticity versus volition in the perceptual has definite opinions of what to do but conforms owing to a change, or the appeal of belonging to a particular group, or even normative influence of the group due to social reasons. In this developmental effects, such as the susceptibility of adolescents model, normative influences are presumed to act through the to peer pressure. aversiveness of being in a minority position (Hornsey et al 2003). To our knowledge, this is the first study of brain activity The amygdala activation in our experiment was perhaps the associated with social conformity and independence. Here, we clearest marker of the emotional load associated with standing present brain imaging findings that provide key biological evi- up for one’s belief. This activation occurred during the period of dence for the major psychological theory of conformity and mental rotation, indicating that the emotional engagement was might help resolve some of the controversy surrounding Asch’s intertwined with the perceptual judgment process. Amygdala conformity effect. Asch’s results have been replicated inconsis- activation has been associated frequently with negative emo- tently (Bond and Smith 1996). The two main questions surround- tional states (LeDoux 2000), but the amygdala is also known to ing this inconsistency are whether conformity is culturally deter- modulate social behavior and to be activated by human faces, mined and whether conformity has changed over time as particularly when the faces have emotional significance (Aharon sociopolitical forces have shaped the relative acceptance of et al 2001; Amaral et al 2003; Singer et al 2004; Whalen et al individualism or collectivism (Larsen 1982; Perrin and Spencer 1998). Because we observed amygdala activation only when 1980). Until now, the only measurements of conformity came participants went against the group and not the computers, the from experimental observation of individuals capitulating to a amygdala activation might have represented the interaction of group and the participants’ self-report. Our results affirm the negative affect with the social salience of information originating hypothesis that brain regions classically associated with percep- from one’s peers—a finding consistent with social norm theory tion can be altered by social influences, although the exact www.sobp.org/journal G.S. Berns et al BIOL PSYCHIATRY 2005;58:245–253 253 mechanism of this modulation has yet to be determined. The Grofman B, Feld SL (1988): Rousseau’s general will: A Condorcetian perspec- flip-side of conformity, independence, was found to be associ- tive. Am Political Sci Rev 82:567–576. ated with subcortical activity changes indicative of emotional Hornsey MJ, Majkut L, Terry DJ, McKimmie BM (2003): On being loud and proud: Non-conformity and counter-conformity to group norms. Br J Soc salience, a finding that lends support to social norm theory Psychol 42:319 –335. (Cialdini and Goldstein 2004). Horvitz JC (2000): Mesolimbocortical and nigrostriatal dopamine responses to salient non-reward events. Neuroscience 96:651– 656. Jordan K, Heinze H-J, Lutz K, Kanowski M, Jancke L (2001): Cortical activa- tions during the mental rotation of different visual objects. Neuroimage All authors participated in the conception and design of this 13:143–152. study. JC, CFZ, GP, and MEM-S were responsible for data Kastner S, Pinsk MA, De Weerd P, Desimone R, Ungerleider LG (1999): In- collection. GSB and JC analyzed data; GSB was the main author creased activity in human visual cortex during directed attention in the of the manuscript, and all authors contributed critical revisions. absence of visual stimulation. Neuron 22:751–761. Supported by an unrestricted charitable contribution for Ladha KK (1992): The Condorcet jury theorem, free speech, and correlated votes. Am J Political Sci 36:617– 634. research in neuroeconomics by Jim Richards. Larsen KS (1982): Cultural conditions and conformity: The Asch effect. Bull Br Supplementary material cited in this article is available Psychol Soc 35:347. online. LeDoux JE (2000): Emotion circuits in the brain.Annu Rev Neurosci 23:155–184. Nobre AC, Coull JT, Maquet P, Frith CD, Vandenberghe R, Mesulam MM Aharon I, Etcoff N, Ariely D, Chabris CF, O’Connor E, Breiter HC (2001): (2004): Orienting attention to locations in perceptual versus mental Beautiful faces have variable reward value: fMRI and behavioral evi- representations. J Cogn Neurosci 16:363–373. dence. Neuron 32:537–551. Perrin S, Spencer CP (1980): The Asch effect—a child of its time? Bull Br Akaike H (1974): A new look at the statistical model identification. IEEE Trans Psychol Soc 32:405– 406. Automatic Control 19:716 –723. Pessoa L, Kastner S, Ungerleider LG (2003): Neuroimaging studies of atten- Alivisatos B, Petrides M (1997): Functional activation of the human brain tion: From modulation of sensory processing to top-down control. J Neu- during mental rotation. Neuropsychologia 35:111–118. rosci 23:3990 –3998. Phan KL, Taylor SF, Welsh RC, Ho S-H, Britton JC, Liberzon I (2004): Neural Amaral DG, Capitanio JP, Jourdain M, Mason WA, Mendoza SP, Prather M correlates of individual ratings of emotional salience: A trial-related fMRI (2003): The amygdala: Is it an essential component of the neural network study. Neuroimage 21:768 –780. for social cognition? Neuropsychologia 41:235–240. Phelps EA, O’Connor KJ, Gatenby JC, Gore JC, Grillon C, Davis M (2001): Arrow KJ (1963): Social Choice and Individual Values, 2nd ed. New Haven, Activation of the left amygdala to a cognitive representation of fear. Nat Connecticut: Yale University Press. Neurosci 4:437– 441. Asch SE (1951): Effects of group pressure upon the modification and distor- Redgrave P, Prescott TJ, Gurney K (1999): The basal ganglia: A vertebrate tion of judgments. In: Guetzkow HS, ed. Groups, Leadership and Men solution to the selection problem? Neuroscience 89:1009 –1023. Research in Human Relations. Pittsburgh: Carnegie Press 177–190. Rees G, Frith CD (1998): How do we select perceptions and actions? Human Asch SE (1952): Social Psychology. New York: Prentice-Hall. brain imaging studies. Phil Trans R Soc Lond B Biol Sci 353:1283–1293. Ashburner J, Friston KJ (1999): Nonlinear spatial normalization using basis Ress D, Heeger DJ (2003): Neuronal correlates of perception in early visual functions. Hum Brain Mapp 7:254 –266. cortex. Nat Neurosci 6:414 – 420. Bond R, Smith PB (1996): Culture and conformity: A meta-analysis of studies Shadlen MN, Newsome WT (2001): Neural basis of perceptual decision in the using Asch’s (1952b, 1956) line judgement task.Psychol Bull 119:111–137. parietal cortex (area LIP) of the rhesus monkey. J Neurophys 86:1916 – Britten KH, Shadlen MN, Newsome WT, Movshon JA (1992): The analysis of 1936. visual motion: A comparison of neuronal and psychophysical perfor- Shepard RN, Metzler J (1971): Mental rotation of three-dimensional objects. mance. J Neurosci 12:4745– 4765. Science 171:701–703. Cialdini RB, Goldstein NJ (2004): Social influence: compliance and confor- Singer T, Kiebel SJ, Winston JS, Dolan RJ, Frith CD (2004): Brain responses to mity. Annu Rev Psychol 55:591– 621. the acquired moral status of faces. Neuron 41:653– 662. Cohen MS, Kosslyn SM, Breiter HC, DiGirolamo GJ, Thompson WL, Anderson AK Stapel DA, Koomen W (1997): Social categorization and perceptual judg- (1996): Changes in cortical activity during mental rotation: A mapping study ment of size: When perception is social. J Pers Soc Psychol 73:1177–1190. using functional magnetic resonance imaging.Brain 119:89 –100. Tagaris GA (1997): Mental rotation studied by functional magnetic reso- Coren S, Enss JT (1993): Size contrast as a function of conceptual similarity nance imaging at high field (4 Tesla): Performance and cortical activa- between test and inducers. Percept Psychophys 54:579 –588. tion. J Cogn Neurosci 9:419 – 432. Duvernoy HM (1999): The Human Brain. Vienna: Springer-Verlag. Whalen PJ, Rauch SL, Etcoff NL, McInerney SC, Lee MB, Jenike MA (1998): Friston KJ, Ashburner J, Poline JB, Frith CD, Heather JD, Frackowiak RSJ Masked presentations of emotional facial expressions modulate amyg- (1995): Spatial registration and normalization of images. Hum Brain dala activity without explicit knowledge. J Neurosci 18:411– 418. Mapp 2:165–189. Zink CF, Pagnoni G, Martin ME, Dhamala M, Berns GS (2003): Human striatal Friston KJ, Holmes AP, Worsely KJ (1999): How many subjects constitute a response to salient nonrewarding stimuli. J Neurosci 23:8092– 8097. study? Neuroimage 10:1–5. Zink CF, Pagnoni G, Martin-Skurski ME, Chappelow JC, Berns GS (2004): Goldstone RL (1995): Effects of categorization on color perception. Psychol Human striatal responses to monetary reward depend on saliency. Neu- Sci 6:298 –304. ron 42:509 –517. www.sobp.org/journal